Catalog Overview
ChimeraMiND is not a single trading bot with a UI wrapped around it. The platform is built from roughly 50 distinct, independently maintained capabilities across nine layers — introspection tooling, execution routes, strategy logic, machine learning, exchange connectivity, and real-time analytics. This page is the index; each linked page below documents the capabilities in that layer with the actual module/file it maps to, so every number here is traceable back to source.
Capability count by layer
| Layer | Count | Detail |
|---|---|---|
| MCP introspection tools | 60 tools / 10 domains | Read-only observability surface (portfolio, analytics, AI state, market data) |
| REST API routes | 57 route files | Trading, portfolio, risk, config, calibration, infra |
| Trading bot types | 20 | DCA, grid, momentum, breakout, cascade, arbitrage variants, and more — see Trading Bots |
| Core strategy modules | 16 | Orchestrated as one ensemble, not run in isolation — see ML Models |
| ML models | 5 | Direction (GBM), regime (HMM), cascade risk, microstructure (DeepLOB), meta-optimizer (RL) |
| Exchange gateways | 7 | Binance, OKX, Bybit, Bitget, Coinbase, Kraken, Paper — see Exchange Gateways |
| Real-time analytics streams | 14 | VPIN, CVD, OI velocity, regime, breadth, and more — see Analytics Streams |
| Core execution subsystems | 24+ | Risk, calibration, portfolio, feeders, liquidation detection, on-chain tracking |
Why the count is presented this way
Each number above is a count of things that actually exist in the codebase — files, registered tools, or configured integrations — not a marketing estimate. The MCP tool count and domain breakdown come directly from chimera_mcp/server.py. The route file count comes from api/routes/. The strategy and model counts come from strategy/ and the model-serving layer. This catalog is meant to survive a technical reviewer opening the repository and checking.
What "antifragile" means here
The platform's core differentiator is not any single model or bot — it is that 16 strategy modules are coordinated through a consensus/voting layer (strategy_mux, decision_engine) into one adaptive decision process, with a reinforcement-learning meta-optimizer continuously adjusting parameters against live performance. See Architecture Overview for how this fits into the runtime.